A Condition Monitoring Method for Wind Turbine Generators Based on Latent Variables
Yao-Ching Yin, Boqiang Xu, Jinglong Wu, Liling Sun · IEEE Transactions on Instrumentation and Measurement · 2025
As the core electrical component of a wind power system, the condition of the generator directly affects the operational stability and power generation efficiency of wind turbines (WTs). However, existing equipment condition monitoring (ECM) methods for WTs struggle to balance comprehensiveness and specificity. Moreover, reliance on a single monitoring indicator makes it challenging to determine whether generator anomalies are caused by external factors, thereby limiting monitoring effectiveness. To address these challenges, this study proposes an ECM strategy that directly monitors latent relationships between the generator and external system features. The strategy integrates latent variable (LV) modeling with an orthogonal nonlinear decoupling mechanism and introduces an explicit dynamic structure to handle the dynamic nonlinear characteristics of WTs for LV extraction. Using the extracted LVs, generator-related and external monitoring indices are constructed to clearly define fault scope and impact. Experimental analysis of various typical fault conditions demonstrates that the proposed method achieves comprehensive generator condition monitoring. Compared with mainstream ECM strategies, this approach exhibits superior robustness and sensitivity.